muzux2 said:
Then why intel killed timna family cpu's??
If you read available articles on Timna, majority of fault to timna's demise was rambus. If Rambus was able to curb the cost, we would have seen amazing stuff coming out of Intel. But that didn't happen, so Intel had to use brigde chip (MTH, or Memory Translation Hub) to link Rambus memory controller with less expensive SDRAM memory. Though that solution worked, due to some serious bugs Intel decided to scrap the entire Timna project.
You can read about Timna here :
Intel Timna - Google Search
Specifically :
PC World - Intel Kills Timna
PCWorld said:
About ten years ago the company created an integrated product line, Brookwood notes. Intel shipped both the 386SL and a related 486SL after some delays. But vendors stayed away, and by 1992 Intel had cancelled the line.
muzux2 said:
With IMC, AMD has trashed all Intel Cpu's in 'memory based benches' say faster mem latency, mem read & write speeds..
AMD's processors were good options against Netburst architecture... they were faster for clock to clock (even sometimes faster than higher clocked Pentiums), they were more energy efficient and competitively priced.
But came Core architecture and started punching Athlons black and blue. No matter what fanboys (like you) said... the fact was in real-world performances Intel burried AMD.
Read following blog article :
Computing Intensive: IMC Myth
To summarize the whole thing...
AMD did not invent IMC nor it was first to come out with processors with IMC.
But it is true that they had successful "run" with IMC.
Intel had 386SL and 486SL available in market and technically it was working IMC solution, but vendors didn't show much interest. Even though Timna was killed before birth, technically the solution was working. It was changed due to business reasons (use of MTH) and later killed due to technical problems in MTH.
As for your memory benchmarks, if you look at bigger picture... excelling in one area didn't matter after all, as Core ruled in overall performance.
-------------
Although bit off-topic... your stance over AMD reminds me of another fanboy rant we kept hearing before Phenom. That, Intel's quad-core solution is just a patch job. AMD's "native" quad-core is better solution, yada yada yada.
I had expressed my views over that long back in one
thread...
let's quote my statement here :
iGo said:
I sometime really don't get the point of this argument about "true" quad-core... what's wrong with two patched dual-cores?
I understand someone might bring up interconnect between two dual-cores are serious bottleneck and so. But the catch here is that the way Intel CPU's work, on FSB. So as long as Intel is providing enough bandwidth via FSB/Cache it doesn't matter. Despite people despising Intel's "patch-work" on quad-core, benchmarks and real-world performance testing has already proved that Intel's quad-core works fine and scales well.
In fact, due to it's IMC AMD can easily put two Phenom dual-cores without worrying about CPU interconnect bandwidth.
While it is true that at some point in future FSB is not going to prove as most effective way to scale bandwidth, Intel already sees that. With Nehalem Intel brings back IMC to it's processors and Direct-Connect architecture (which they call QuickPath).
What you need to understand is, with so-called patching two Dual-Cores Intel can scale well with it's superior production setup and can introduce newer products faster than competition's. While we still await AMD's answer to Core (so far Phenom's early previews look like competitive product than knock-out punch to Core), Intel is progressing faster to better die-shrinks and able to deliver as per their roadmap (even bit earlier than promised, look at penryn).
So when Phenom launches, it has to pit against 45nm Intel CPUs with it's 65nm die. It might stand eye-to-eye to Intel's offering in performance, but Intel already has almost a lead of Year. Intel has made it self good name with enthusiast & gaming community and you don't need to be geek to understand that 45nm with proven & mature architecture (with newer tweaks) will give Intel edge over new architecture on 65nm by AMD.
For example, just look at the TDP on Phenom 9700 : 2.6GHz - 125W
and here's TDP on similarly clocked Yorkfield Q9450 : 2.66 GHz - 95W
Sorry for my rambling... but I really find the wholeTrue Quad-Core argument pointless unless numbers can prove it wrong in real-world.
But Phenom came and didn't stir things up. In fact, because of the overall talk of "true" and "native" quad-core by fanboys and AMD itself... the performance of Phenom (the very first batch) was sort of embarrassment. Today, after AMD started cleaning up the mess and fixing things Phenom (not to mention some aggressive price cuts), Phenom is at least considerable alternative to Core processors. Again, lion's share for that goes to AMD's spider platform.
-------------
Just to make it clear to you. I am NOT Intel fanboy. I've not used AMD chip ever in my life, but that wasn't because I am fan of Intel.
I want AMD to stay in game as much as any AMD fanboy would want. Because it's always good to have alternative and give each company a reason to innovate.
But I prefer to see things rationally and not following something blindly. I had used crap called Pentium D for sometime, and believe me I almost made a switch to AMD... had it not been Core's launch. Like the popular saying goes... "Ugate Suraj ko sab salaam karte hai" (Everyone salutes rising sun) !!